Kappa results from altered datasets

Preliminary steps

Load libraries

Load files

Obtain the aggregated curves from data

Mean KLC curves for each level of noise

df1 contains the mean kappa loss value of all datasets for each level of noise only.

Plot the results

Plot each technique separately

Plot all techniques overlapping

Plot as a grid

Mean KLC curves for each level of noise and quartile of instances

df2 contains the mean kappa loss value of all datasets for each decile and level of noise.

Plot the results

Plot each technique separately

Plot all techniques overlapping

for(instance in instances_names) {
  # Filter data for the current instance percentage
  filtered_data <- subset(df2, percentage == instance)
  
  # Create plot
  p2 <- ggplot(filtered_data, aes(x = noise, y = kappa_loss, color = factor(technique))) +
  geom_point() +
  geom_line(aes(noise)) +
  labs(x = "Noise", y = "Kappa Loss", color = "Technique") +
  ggtitle(paste0("Kappa Loss Curves by technique, noise and ", instance, " % of instances altered")) +
  theme_bw() +
  scale_y_continuous(limits = c(0.0, 0.5), breaks = seq(0, 1, by = 0.1))
  
  # Print plot
  print(p2)
}

Plot as grid for deciles

# Create plot
p2 <- ggplot(df2, aes(x = percentage, y = kappa_loss, color = factor(noise))) +
  geom_point() +
  geom_line(aes(percentage)) +
  labs(x = "Noise", y = "Kappa Loss") +
  ggtitle("Kappa Loss Curves by technique, noise and percentage of instances altered") +
  theme_bw() +
  scale_y_continuous(limits = c(0.0, 0.5), breaks = seq(0, 1, by = 0.1)) +
  facet_wrap(~ technique)

# Print plot
print(p2)

ggsave("results/plots/KLC_means_instances.png", p2, width = 20, height = 16, dpi = 600)

Plot as grid for quartiles

# Create plot
p3 <- ggplot(df2_q, aes(x = percentage, y = kappa_loss, color = factor(noise))) +
  geom_point() +
  geom_line(aes(percentage)) +
  labs(x = "Noise", y = "Kappa Loss") +
  ggtitle("Kappa Loss Curves by technique, noise and percentage of instances altered") +
  theme_bw() +
  scale_y_continuous(limits = c(0.0, 0.5), breaks = seq(0, 1, by = 0.1)) +
  facet_wrap(~ technique)

# Print plot
print(p3)

ggsave("results/plots/KLC_means_instances_q.png", p3, width = 20, height = 16, dpi = 600)